Bryan Wilder Field Lozano
Bryan Wilder Field Lozano
> Also, I ended up implementing my own SIMD version to mass convert float into bfloat16. I am not sure how often TT needs to convert large amounts of FP32...
This should be possible now! Thankyou @broskoTT https://github.com/tenstorrent/tt-umd/commit/ddc61ff7941dc27522647927811d22fc25016efe
See #13785 for export of MEM_MAP_END
@pgkeller so the NOC_0* macros should become Hal class methods, correct?
> > [@pgkeller](https://github.com/pgkeller) so the NOC_0* macros should become Hal class methods, correct? > > funny, those macros are misnamed (NOC_0 but take noc_index as a parameter). yes, should be...
@abhullar-tt sorry, I accidentally closed this. It is not completely done, but there are no more components here that block ARCH_NAME removal. I think the file is only included behind...
Looks like I will have to circle back and fix debug build: https://github.com/tenstorrent/tt-metal/actions/runs/14242342258/job/39915958243 Should be easy fix though.
@ayerofieiev-tt what is it you don't really love about this PR? 1. Too many forward declarations? (I would be okay to not do this in this first pass). 2. Too...
This would be especially helpful, when people use type aliases, and you can't find the template instantiations in the code.
@afuller-TT could you please keep the contents of this PR in mind as you work on packaging stuff? I really like the style I've used here for MetaliumConfig.cmake.in and the...